WeLeBe (We Learn Better)

architecture
code
result. video to text
lectures

💡 Inspiration

Despite the epidemic, online conferencing has played a critical role in helping educational institutions to hold lessons. It does, however, have certain disadvantages. Students find it difficult to concentrate attention for lengthy periods of time in a virtual environment, which causes them to lose out on vital information.

We aim to aid the students in this aspect, to ensure they do not miss out on anything.

❓ What it does

*Processes the videos, providing detailed summaries and notes on subjects presented in class, ensuring that students stay on track. An automatic interface with DropBox is also included, allowing students to access the content anytime they choose.

*The automatic question generator, which is driven by a transformer Neural Network architecture, leverages the notes it creates to suggest possible test questions based on the topic presented in that session, as well as the perfect solutions to the created questions. We feel that this functionality adds significant value and, as a result, improves the learning experience.

*The ease with which it can be used. The user only needs to upload the movie to our service, and SMLR will handle everything else.

*Dashboard to organize the processed lectures.

*After processing, the user will get push alerts.

🏃‍♂️ Challenges we ran into

*The time constraint: Integrating many features into the program in a single day proved difficult, and there was limited time for testing. It was, nevertheless, a valuable learning experience.

*We experienced several difficulties connecting Video to Text with the backend and had to rely on free services.

*We had to cycle over the text and run it numerous times to get the entire summary because the T5 transformer can only accept 512 tokens. When we divide, we'll also lose some important situations.